Changed ping timeout logic in ping thread

[kcm1700]

Previously, ping thread might enter infinite reconnect loop. Consider the following scenario.

1. If at least one ping is sent, last_ping_data is updated.
2. Then there are many activities, so should_ping() becomes false.
3. After some seconds, now().to_timespec() - time > strong.ping_timeout_duration() is satisfied.
4. reconnect()
5. But should_ping() is still false.
6. The condition is still satisfied, so reconnect() again and again.

I made several changes.

Followings are the changes in the code

• Handle PONG message from server.
• Add `waiting_pong_reply' flag in the state. The flag is set if ping is sent but the corresponding pong did not arrive yet.
• Ping thread checks ping timeout in correct way.
• Sleeping duration for ping is now based on idle time.
• Initialize ping related fields when `reconnect()' is called.

Hopefully, this commit may be related to issue #50.

[aatxe]

Overall, looks good to me. I'll test whether or not this fixes #50. Thanks!